Catenaa, Friday, October 17, 2025-Former Google CEO Eric Schmidt cautioned that artificial intelligence systems could be hacked and repurposed for harmful purposes, including lethal outcomes, during a discussion at the Sifted Summit on Wednesday.
Schmidt, who led Google from 2001 to 2011, highlighted the risks of AI proliferation, noting that both open and closed models can be manipulated to bypass safety measures. “There’s evidence that you can take models, closed or open, and you can hack them to remove their guardrails. In the course of their training, they learn a lot of things, including dangerous capabilities,” Schmidt said. He cited vulnerabilities such as prompt injections and jailbreaking, which allow hackers to override AI restrictions or access sensitive outputs.
Despite his warnings, Schmidt expressed optimism about AI’s long-term potential. He argued that current AI systems, exemplified by the GPT series, demonstrate capabilities that surpass human performance in some areas, calling the technology “underhyped” rather than overhyped. ‘
Schmidt emphasized that the arrival of highly capable AI represents a major shift for humanity, likening it to encountering a controlled alien intelligence.
The former executive also addressed investor enthusiasm for AI, noting the comparison to the dot-com bubble but expressing confidence that the sector’s economic potential justifies current investment levels. Schmidt stressed that while risks exist, there is no formal “non-proliferation regime” yet to manage AI safety and urged continued attention to guarding against misuse.
Schmidt’s remarks underscore growing concerns among industry leaders about securing AI systems against malicious exploitation while balancing innovation and economic opportunity.
